# [[Message about Meeting with Enrico]] #llmchat #paper/generatingphilosophy Attached to two things. One is quite a lengthy conversation about a meeting I had. The other is the paper that the meeting is talking about. I would like you to consider both documents and I would like you to come up with a strategy of things I should research or get clear in my mind so as to work towards the goals that were set within the meeting or the ideas that were suggested in the meeting. IMPORTANT: Please remember to follow your [[system prompt]] to the letter!! ## draft Draft/Plan 20 May 2025 #draft #paper/generatingphilosophy 1. Introduction “Forty-two," said [[Deep Thought]], with infinite majesty and calm. It was a long time before anyone spoke. Out of the corner of his eye Phouchg could see the sea of tense expectant faces down in the square outside. "We're going to get lynched aren't we?" he whispered. "It was a tough assignment," said [[Deep Thought]] mildly. "Forty-two!" yelled Loonquawl. "Is that all you've got to show for seven and a half million years' work?" "I checked it very thoroughly," said [[the computer]], "and that quite definitely is the answer. I think [[the problem]], to be quite honest with you, is that you've never actually known what [[the question]] is." -- [[Douglas Adams]], [[The Hitchhiker]]'s Guide to the Galaxy In [[The Hitchhiker]]'s Guide to the Galaxy, humanity asks an AI to do some philosophy. An computer named [[Deep Thought]] is constructed and instructed to provide "The Answer to the [[Ultimate Question]] of Life, the Universe, and Everything" (REF). Humanity builds this computer only to receive the answer '42'—an answer which, while apparently correct (at least according to [[Deep Thought]]), means next to nothing at all due to humanity’s failure to [[know what]] the [[Ultimate Question]] in fact is. In 2025 we are in a position to think about the relationship between philosophy and AI for real. Can LLMs enhance philosophical understanding? They are happy to dispense philosophical wisdom if we ask them to, but should we listen? I suspect many would doubt that we should, notoriously prone to 'hallucinations', and the banal, hyperbolic essays churned out by ChatGPT have become the scourge of undergraduate teaching. In this paper, I argue for cautious optimism: the creators of such systems sometimes describe them as reasoning engines, and I suggest this description is broadly accurate. Even if LLMs do not truly reason, their ability to simulate human reasoning enables them to enhance the philosophical understanding of their users Can an LLM generate good quality philosophy? good quality = roughly, publishable in a good quality journal The paper argues that an LLM could achieve this with minimal architectural changes from existing reasoning systems. The main adaptation would be a new evaluation loop designed to reward philosophical virtues. To develop this argument, the paper will: Define "good quality philosophy" through abduction and theoretical virtues. Analyse DeepSeek-Prover-V2 as an example of advanced AI reasoning. Introduce 'PhiloSeeker,' a model for AI philosophy, covering its design, training, evaluation, and potential objections. Conclude with a summary of architectural needs. The overall aim is to explore how LLMs might be trained to produce high-quality philosophy by targeting these virtues. 2. Clarifying the Scope: Producing vs. Doing Philosophy This paper focuses on whether an LLM can produce outputs that constitute good quality philosophy. This question is distinct from, and less demanding than, asking whether LLMs can 'do philosophy' in the way humans do, whether they can think, understand, or perform genuine speech acts like assertion. The idea that LLM outputs can increase understanding, or be philosophically valuable, might be met with the objection that these systems do not themselves understand anything. They might be seen as "stochastic parrots," merely arranging words in statistically plausible patterns. While current LLM architecture differs significantly from the human mind, we should not assume that human-like reasoning or understanding is a prerequisite for producing philosophically illuminating outputs. Drawing on Butlin and Viebahn, it is plausible that fine-tuned LLMs can produce outputs with a "descriptive function." This is defined as "the function of conveying information to an observer or consumer system, so as to cause the consumer to behave as though some condition holds." Pre-trained LLMs primarily predict likely word sequences. Fine-tuning, however, involves additional training aimed at producing outputs that are accurate, relevant, and informative, often by rewarding outputs consistent with external sources or positive human evaluations. This fine-tuning process can equip LLM outputs with a descriptive function, making them reliable enough to help users refine their understanding of the world, analogous to how a thermometer is designed to indicate temperature. Therefore, even if an LLM itself lacks understanding and is not performing speech acts, if its outputs are a product of fine-tuning for accuracy and informativeness (i.e., they have a descriptive function), they might provide correct or insightful philosophical claims. Such outputs, if they possess the relevant theoretical virtues, could be considered "good quality philosophy" in terms of their content and structure, regardless of the internal states or processes of the LLM. This paper examines the potential for producing such outputs. 3. Conceptual Foundations: Abduction and the Systematisation of Theoretical Virtues This section establishes the philosophical framework for the paper. Abductive Reasoning (following Timothy Williamson): Defined as inference to the best explanation. Williamson argues for its importance not only in natural sciences but also in mathematics and philosophy. Theories are chosen not simply for logical entailment of evidence, but for their overall explanatory ability. An example from mathematics: Williamson (REF) argues that even basic mathematical principles (e.g., axioms of set theory like ZFC) are supported by abductive reasoning. ZFC is accepted because it successfully and powerfully organises mathematical findings elegantly, not because its axioms are all self-evidently proven from more basic steps. This supports the use of abduction in "armchair" disciplines like philosophy, suggesting they share a core method with empirical fields. Philosophical theories are seen as attempts to offer the best explanation for observations, conceptual differences, or intuitions. Philosophical model-building (Williamson 2017a) is presented as a clear example of abduction in action, where preference for simpler/elegant models is justified abductively for clearer explanations and avoiding over-fitting. Keas's (2018) Systematisation of Theoretical Virtues: While Williamson identifies overall explanatory ability as key, his account does not focus extensively on a systematic breakdown of what constitutes such ability. He does mention features such as "simplicity, elegance, strength – meaning how informative and general a theory is – and its power to unify different ideas." However, a more detailed and organised understanding of what makes an explanation "best" or "virtuous" is needed for designing an LLM. Williamson's account can be supplemented by Keas's research, which offers a detailed classification of theoretical virtues. This provides a richer vocabulary for what an abductive philosophical inquiry (or an LLM emulating it) might aim for. Keas categorises numerous theoretical virtues into four main groups: Evidential Virtues: Concerning how well a theory explains evidence (e.g., explanatory depth, evidential accuracy). Coherential Virtues: Relating to a theory's internal logical soundness and its fit with other knowledge (e.g., internal coherence, universal coherence). Aesthetic Virtues: Pertaining to qualities like elegance and simplicity (e.g., simplicity, unification). Diachronic Virtues: Concerning how a theory performs and changes over time (e.g., durability, fruitfulness). This structured taxonomy offers a comprehensive set of standards for evaluating theories. The paper argues that Williamson’s abduction and Keas’s classification are mutually supportive. Keas articulates the standards implicit in Williamson's "best" explanation. For instance, Williamson's "simplicity" and "elegance" map to Keas's aesthetic virtues; Williamson's "strength" or "unificatory capacity" are expanded by Keas's unification and aspects of explanatory depth. This combined understanding (philosophical thinking as abductive activity, aiming for virtues detailed by Keas) is posited as something that can, in principle, be implemented in an LLM. 4. The Architecture of DeepSeek-Prover-V2 (DSP-v2): An Analogous Case Study To assess the feasibility of an LLM generating good quality philosophy, this section introduces DeepSeek-Prover-V2 (DSP-v2). DSP-v2 is not presented merely as a recent LLM advancement, but as an analogous case study for several specific reasons: Demonstrating Capacity for Complex Reasoning: DSP-v2's success in formal mathematical theorem proving shows that current LLMs can perform complex, goal-directed, and highly structured reasoning. This establishes a baseline capability relevant to the demands of philosophical argumentation. Illustrating Operationalisable Abduction-like Processes: The way DSP-v2 generates high-level proof sketches and then decomposes these into verifiable sub-steps provides a model for how abductive-like reasoning (forming a hypothesis and then testing its components) can be implemented in an LLM. Showing Implicit "Virtue-Seeking" in a Formal Domain: DSP-v2's training, particularly its use of a consistency reward, implicitly guides it towards outputs that exhibit formal analogues of theoretical virtues, such as internal coherence with an initial plan and logical consistency. This suggests that LLMs can be oriented towards desirable output characteristics beyond mere statistical likelihood. Providing an Architectural and Conceptual Precedent: The specific architecture of DSP-v2 (e.g., hierarchical decomposition, feedback loops based on verification) offers a concrete starting point for conceptualising PhiloSeeker. It makes the proposal for adapting such principles to the less formal domain of philosophy appear as a plausible extension of existing AI methodologies. DSP-v2 Overview: An open-source LLM for formal theorem proving in Lean 4 (Ren et al. 2025). Objective: Generate formal mathematical proofs, linking informal intuition to formal verification. Core Strategy: Hierarchical decomposition, guided by informal reasoning. A general LLM (DeepSeek-V3) generates a natural language ‘proof sketch’. This sketch acts as a blueprint for decomposing the problem into formal subgoals (lemmas). Proof Generation Process: The sketch is translated into formal Lean 4 statements (e.g., have clauses with sorry placeholders). Subgoals are tackled individually by a specialised prover model (e.g., a 7B parameter model mentioned in their research). Successful sub-proofs are synthesised into a complete, verified proof. Learning and Refinement Mechanisms: ‘Cold-start’ reasoning data: Pairs successful informal sketches with their verified formal proofs for training. Reinforcement Learning (RL): Primary reward: Binary (correct/incorrect Lean verification). ‘Consistency reward’: Encourages structural alignment between the final proof and the initial sketch. Expert Iteration: Incorporates successfully generated and verified proofs into the supervised fine-tuning (SFT) dataset. Operational Modes: ‘non-CoT’ (non-Chain-of-Thought) mode: For rapid, concise formal proof generation. ‘CoT’ (Chain-of-Thought) mode: Articulates intermediate reasoning steps in natural language for transparency. These architectural and learning features of DSP-v2 provide a concrete foundation for considering how similar principles might be adapted for the distinct challenges of philosophical reasoning, as explored in the following sections. 5. DSP-v2: Operationalising Abduction and Implicitly Seeking Theoretical Virtues This section analyses DSP-v2's operational processes, as described in Section 4, through the conceptual lens of abduction and theoretical virtues (from Section 3). The aim is to demonstrate how its mechanisms can be interpreted as formal analogues of these philosophical concepts. DSP-v2's Core Methodology as an Analogue to Abductive Process: The generation of an initial informal proof sketch by DeepSeek-V3 is analogous to the abductive step of formulating a high-level hypothesis or potential explanation (i.e., "This theorem is true, and this sequence of major steps offers a plausible way to demonstrate its truth"). The subsequent decomposition of this sketch into formal subgoals, and the systematic attempt to prove each one, mirrors the elaboration and rigorous testing of an initial abductive hypothesis. Each successfully proven subgoal lends support to the overarching sketch. A fully verified proof, therefore, confirms the initial sketch as a "good" explanation or a successful strategic outline within the formal system. Through its training (RL and expert iteration), DSP-v2 learns to generate sketches that are more likely to lead to successful proofs, akin to an abductive reasoner learning to propose better, more tractable hypotheses. DSP-v2's Architecture and Learning Mechanisms as Implicitly Seeking Formal Analogues of Theoretical Virtues: Evidential Accuracy and Internal Consistency (Formal Analogues): The Lean 4 proof assistant acts as an objective arbiter. Any proof verified by DSP-v2 is, by definition, formally correct (analogous to ‘evidential accuracy’ relative to the axioms of the formal system) and internally consistent (containing no detectable logical contradictions). These are non-negotiable requirements of its operational environment. Internal Coherence (Formal Analogue): The ‘consistency reward’ in DSP-v2’s RL phase directly promotes a formal counterpart to this virtue. By rewarding proofs that maintain structural alignment with the initial lemma decomposition, the system is encouraged to produce outputs where sub-proofs fit together cohesively and follow the strategic plan. This is analogous to internal coherence, which values the integrated and non-ad hoc arrangement of a theory's components. Simplicity (Formal Analogue): The ‘non-CoT’ mode, optimised for concise formal proofs, aims for a form of syntactic or structural simplicity—achieving the deductive goal with less complex formal output. This parallels the theoretical virtue of simplicity. Clarity and Explanatory Depth (Formal Analogues): DSP-v2’s ‘CoT’ mode, by explicitly articulating intermediate reasoning steps in natural language alongside formal code, enhances transparency and understandability. This provides a form of ‘explanatory depth’ by revealing the inferential structure and rationale, analogous to the philosophical desire for clear and insightful explanations. Learning from "Virtuous" Examples: Through its training on "cold-start" data (its own successfully verified proofs, which are inherently accurate, consistent, and often coherent due to the consistency reward), DSP-v2 learns from exemplars that already possess these desirable formal characteristics. In essence, while DSP-v2 does not engage in philosophical deliberation, its design illustrates how an LLM can be structured to navigate a complex reasoning space using abduction-like strategies and be guided by feedback mechanisms that implicitly select for outputs possessing characteristics analogous to theoretical virtues. This analysis provides the crucial groundwork for proposing that similar principles can be adapted for the philosophical domain in the PhiloSeeker model. 6. PhiloSeeker: A Hypothetical Adaptation of DSP-v2 for Virtuous Philosophy This section proposes 'PhiloSeeker,' a hypothetical LLM adapted from DSP-v2's principles to aim for theoretically virtuous philosophical outputs. Objective and Domain: Primary Objective: Generate philosophical texts (theories, arguments, conceptual analyses, responses to problems) that are cogent, internally consistent, and exhibit high theoretical virtue (as per Keas, 2018). Operational Domain: Natural language philosophical discourse, with its nuances and lack of formal verifiability (unlike Lean 4). Core Strategy (adapting DSP-v2's hierarchical decomposition): A sophisticated general LLM, ‘PhiloSketcher’ (analogous to DeepSeek-V3), would produce an initial high-level ‘theory sketch’ or ‘argumentative blueprint’ in response to a philosophical problem/prompt. This blueprint would articulate main claims, define key concepts, outline sub-argument structures, and potentially flag objections or areas needing clarification, decomposing the task. Process of Philosophical Construction (mirroring DSP-v2's sketch-to-subgoal translation): ‘PhiloSketcher’ translates its blueprint into a structured outline (assertions, questions, tasks with placeholders for elaboration). These individual philosophical sub-tasks (e.g., "provide three arguments for claim X," "define 'phenomenal consciousness'," "formulate response to Gettier problem") would be processed by a more specialised ‘Philosophical Argumentation Model’ (analogous to DSP-v2’s smaller prover model). Adapted Learning and Refinement Mechanisms: ‘Synthetic data for a philosophical cold start’: When PhiloSeeker successfully constructs a complete philosophical theory/argument deemed highly virtuous by evaluators (human or AI), this entire output is paired with the initial argumentative blueprint. This creates training examples linking high-level philosophical strategy with detailed, virtuous execution in natural language. Reinforcement Learning (RL): Central to development, but with a different reward structure than DSP-v2's binary reward. Guided by a composite ‘virtue score’ derived from evaluating the text against Keas’s theoretical virtues (via automated metrics and human feedback, detailed in §7). A critical component: a ‘blueprint-consistency reward’ (paralleling DSP-v2) to incentivise the output to faithfully instantiate the initial sketch, promoting internal coherence. Expert Iteration (adapted): PhiloSeeker learns from its own most ‘virtuous’ outputs (those receiving high ratings from human evaluators or a virtue-discriminator model). These are incorporated back into its training set. Dual Generation Modes (similar to DSP-v2): ‘non-CoT’ mode: Might produce concise, polished philosophical essays or arguments. ‘CoT’ mode: Would explicitly articulate the underlying argumentative blueprint, rationale, definitions, argument development, counter-argument consideration, and connections between theory parts, enhancing transparency and scrutiny. The section suggests DSP-v2's core architectural loop (sketch, decompose, solve, synthesise, refine) is robustly applicable to philosophy. The primary adaptation lies in the nature of the ‘solved state’ (philosophically virtuous output) and the evaluation/reward mechanisms. 7. Training and Evaluating PhiloSeeker This section details methods for training and evaluating PhiloSeeker, emphasising operationalising theoretical virtues (Keas 2018) as the guiding principle. Core Training Loop Adaptations from DSP-v2: Initial Data Seeding: A substantial corpus of high-quality philosophical texts, selected for philosophical significance and exemplification of virtues (clarity, rigour, coherence, etc.). Corpus might require annotation (human or LLM-driven) to highlight argumentative structures, definitions, or virtue instantiations for initial supervised fine-tuning of ‘PhiloSketcher’ and ‘Philosophical Argumentation Model’. ‘Cold-start’ Data Generation (Philosophical): When PhiloSeeker produces a complete philosophical work (essay, theory fragment) from a blueprint, and it is deemed highly virtuous by the evaluation framework, this blueprint-to-output sequence becomes a training instance. Curriculum Learning: ‘PhiloSketcher’ could generate numerous smaller, focused philosophical tasks (e.g., "define 'free will' compatibly with determinism," "formulate three objections to utilitarianism"). These provide denser, targeted training signals for the ‘Philosophical Argumentation Model’. Expert Iteration (Philosopher-in-the-loop refinement): PhiloSeeker learns from its own generated philosophical outputs that achieve a high ‘virtue score’ (determined by the evaluation process). These are added back to the training set. Reinforcement Learning (RL) with a ‘Virtue Score’: This is the most critical adaptation from DSP-v2's binary reward. The score is multi-faceted, derived from Keas’s taxonomy. Operationalising Evidential Virtues: Rewards for consistency with specified empirical data/scientific theories (if prompt demands), or accurate representation/citation of texts. Judging ‘causal adequacy’ or ‘explanatory depth’ of novel claims would likely initially require significant human feedback to train a discriminator. Operationalising Coherential Virtues: NLP tools for detecting overt logical contradictions or inconsistent terminology (for ‘internal consistency’). The ‘blueprint-alignment reward’ strongly promotes ‘internal coherence’. ‘Universal coherence’ could be partially checked against a curated knowledge base. Operationalising Aesthetic Virtues: ‘Beauty’ would likely rely on human ratings to train a discriminator. ‘Simplicity’ might use proxy metrics (ontological parsimony via entity counts; conceptual parsimony via undefined primitive terms; syntactic simplicity via conciseness relative to scope). ‘Unification’ rewarded for connecting disparate problems or deriving many consequences from few principles. Operationalising Diachronic Virtues: ‘Durability’ tested by subjecting theories to ‘objection-bot’ LLMs. ‘Fruitfulness’ might be rewarded for theories identifying unresolved questions or suggesting novel lines of inquiry. Balancing these virtues would likely require multi-objective RL with careful weighting. Hybrid Evaluation Framework (in absence of formal checker): Automated proxy metrics: First-pass signal for certain virtues (consistency, some simplicity forms). Human expert evaluation: Indispensable for qualitative virtues (explanatory depth, beauty, fruitfulness, plausibility). Panels of philosophers with diverse specialisations rating outputs on virtue dimensions using clear rubrics and inter-rater reliability checks. RLHF (Reinforcement Learning from Human Feedback) techniques: Iterative sampling and focused feedback. ‘Virtue-discriminator’ LLM: Trained on human ratings to serve as a scalable proxy for direct human judgment, guiding RL and flagging outputs for human review, with periodic recalibration. 8. Potential Objections and Pluralism This section considers potential objections to the PhiloSeeker project, as outlined in the original draft. Objection: Need for a unique human 'creative spark': The paper's proposal of a system generating complex, virtuous philosophical content implicitly challenges this. Objection: Risk of homogenising philosophical thought: The draft does not detail specific safeguards in this section, but the broader emphasis on diverse virtues and training data might be relevant. Objection: Subjectivity of aesthetic virtues: The paper proposes addressing this by using human ratings to train discriminator models for these virtues. Safeguard Mentioned: Universal coherence with empirical science: This suggests a grounding mechanism to ensure generated theories align with established knowledge where applicable. 9. Summary of Architectural Adaptations This section summarises the key architectural changes proposed for transitioning an LLM from mathematical proof generation to philosophical theory generation. Core LLM Architecture Largely Retained: GPT-style models. Hierarchical prompting strategies. Reinforcement learning (RL) infrastructure. Critical Swap: The Lean-verification filter (used in DSP-v2) is replaced with a 'philosophical virtue-evaluation module.' Additions for the New Evaluation Module: A virtue-discriminator LLM (trained on human judgments of philosophical virtue). A human rating pipeline. Potentially 'objection-bots' to test theory durability. Retuned RL Rewards: Shift from binary correctness to multi-objective optimisation for Keas's theoretical virtues. Maintenance of a blueprint-consistency reward to ensure coherence with the initial plan. 10. Conclusion The paper reiterates its main argument: few fundamental architectural changes are needed to transition from mathematical proof generation to philosophical theory generation. Primary Adaptation: The replacement of a formal verifier with a complex system for evaluating philosophical virtue. This evaluation system is central to the PhiloSeeker concept. Challenges of Philosophical Language: The paper acknowledges that the nuances, ambiguities, and lack of formal verifiability in philosophical language present challenges. However, these are framed as primarily affecting training data creation and evaluation mechanisms, not necessarily posing insurmountable obstacles for the core LLM architecture. Prospect of 'PhiloSeeker': Presented as a conceptually coherent and challenging, yet potentially near-term, research project. Broader Contribution: The endeavour to build such a system would not only advance AI capabilities but also contribute to a deeper and more operational understanding of philosophical virtue itself, by necessitating the clear articulation and implementation of criteria for what constitutes good philosophical work. ## transcript ** It's not a paper, it's still halfway between a draught and a plan. I mean, it's struck there's a lot of stuff in there and it makes sense. So, anyway, read more on yourself. That's one of these and maybe then have another one with more. The other, I want to know what you think of the ideas first, because it's very different arguments. # Message about Meeting with Enrico #llmchat #paper/generatingphilosophy Attached to two things. One is quite a lengthy conversation about a meeting I had. The other is the paper that the meeting is talking about. I would like you to consider both documents and I would like you to come up with a strategy of things I should research or get clear in my mind so as to work towards the goals that were set within the meeting or the ideas that were suggested in the meeting. IMPORTANT: Please remember to follow your system prompt to the letter!! ## draft Draft/Plan 20 May 2025 #draft #paper/generatingphilosophy 1. Introduction “Forty-two," said Deep Thought, with infinite majesty and calm. It was a long time before anyone spoke. Out of the corner of his eye Phouchg could see the sea of tense expectant faces down in the square outside. "We're going to get lynched aren't we?" he whispered. "It was a tough assignment," said Deep Thought mildly. "Forty-two!" yelled Loonquawl. "Is that all you've got to show for seven and a half million years' work?" "I checked it very thoroughly," said the computer, "and that quite definitely is the answer. I think the problem, to be quite honest with you, is that you've never actually known what the question is." -- Douglas Adams, The Hitchhiker's Guide to the Galaxy In The Hitchhiker's Guide to the Galaxy, humanity asks an AI to do some philosophy. An computer named Deep Thought is constructed and instructed to provide "The Answer to the Ultimate Question of Life, the Universe, and Everything" (REF). Humanity builds this computer only to receive the answer '42'—an answer which, while apparently correct (at least according to Deep Thought), means next to nothing at all due to humanity’s failure to know what the Ultimate Question in fact is. In 2025 we are in a position to think about the relationship between philosophy and AI for real. Can LLMs enhance philosophical understanding? They are happy to dispense philosophical wisdom if we ask them to, but should we listen? I suspect many would doubt that we should, notoriously prone to 'hallucinations', and the banal, hyperbolic essays churned out by ChatGPT have become the scourge of undergraduate teaching. In this paper, I argue for cautious optimism: the creators of such systems sometimes describe them as reasoning engines, and I suggest this description is broadly accurate. Even if LLMs do not truly reason, their ability to simulate human reasoning enables them to enhance the philosophical understanding of their users Can an LLM generate good quality philosophy? good quality = roughly, publishable in a good quality journal The paper argues that an LLM could achieve this with minimal architectural changes from existing reasoning systems. The main adaptation would be a new evaluation loop designed to reward philosophical virtues. To develop this argument, the paper will: Define "good quality philosophy" through abduction and theoretical virtues. Analyse DeepSeek-Prover-V2 as an example of advanced AI reasoning. Introduce 'PhiloSeeker,' a model for AI philosophy, covering its design, training, evaluation, and potential objections. Conclude with a summary of architectural needs. The overall aim is to explore how LLMs might be trained to produce high-quality philosophy by targeting these virtues. 2. Clarifying the Scope: Producing vs. Doing Philosophy This paper focuses on whether an LLM can produce outputs that constitute good quality philosophy. This question is distinct from, and less demanding than, asking whether LLMs can 'do philosophy' in the way humans do, whether they can think, understand, or perform genuine speech acts like assertion. The idea that LLM outputs can increase understanding, or be philosophically valuable, might be met with the objection that these systems do not themselves understand anything. They might be seen as "stochastic parrots," merely arranging words in statistically plausible patterns. While current LLM architecture differs significantly from the human mind, we should not assume that human-like reasoning or understanding is a prerequisite for producing philosophically illuminating outputs. Drawing on Butlin and Viebahn, it is plausible that fine-tuned LLMs can produce outputs with a "descriptive function." This is defined as "the function of conveying information to an observer or consumer system, so as to cause the consumer to behave as though some condition holds." Pre-trained LLMs primarily predict likely word sequences. Fine-tuning, however, involves additional training aimed at producing outputs that are accurate, relevant, and informative, often by rewarding outputs consistent with external sources or positive human evaluations. This fine-tuning process can equip LLM outputs with a descriptive function, making them reliable enough to help users refine their understanding of the world, analogous to how a thermometer is designed to indicate temperature. Therefore, even if an LLM itself lacks understanding and is not performing speech acts, if its outputs are a product of fine-tuning for accuracy and informativeness (i.e., they have a descriptive function), they might provide correct or insightful philosophical claims. Such outputs, if they possess the relevant theoretical virtues, could be considered "good quality philosophy" in terms of their content and structure, regardless of the internal states or processes of the LLM. This paper examines the potential for producing such outputs. 3. Conceptual Foundations: Abduction and the Systematisation of Theoretical Virtues This section establishes the philosophical framework for the paper. Abductive Reasoning (following Timothy Williamson): Defined as inference to the best explanation. Williamson argues for its importance not only in natural sciences but also in mathematics and philosophy. Theories are chosen not simply for logical entailment of evidence, but for their overall explanatory ability. An example from mathematics: Williamson (REF) argues that even basic mathematical principles (e.g., axioms of set theory like ZFC) are supported by abductive reasoning. ZFC is accepted because it successfully and powerfully organises mathematical findings elegantly, not because its axioms are all self-evidently proven from more basic steps. This supports the use of abduction in "armchair" disciplines like philosophy, suggesting they share a core method with empirical fields. Philosophical theories are seen as attempts to offer the best explanation for observations, conceptual differences, or intuitions. Philosophical model-building (Williamson 2017a) is presented as a clear example of abduction in action, where preference for simpler/elegant models is justified abductively for clearer explanations and avoiding over-fitting. Keas's (2018) Systematisation of Theoretical Virtues: While Williamson identifies overall explanatory ability as key, his account does not focus extensively on a systematic breakdown of what constitutes such ability. He does mention features such as "simplicity, elegance, strength – meaning how informative and general a theory is – and its power to unify different ideas." However, a more detailed and organised understanding of what makes an explanation "best" or "virtuous" is needed for designing an LLM. Williamson's account can be supplemented by Keas's research, which offers a detailed classification of theoretical virtues. This provides a richer vocabulary for what an abductive philosophical inquiry (or an LLM emulating it) might aim for. Keas categorises numerous theoretical virtues into four main groups: Evidential Virtues: Concerning how well a theory explains evidence (e.g., explanatory depth, evidential accuracy). Coherential Virtues: Relating to a theory's internal logical soundness and its fit with other knowledge (e.g., internal coherence, universal coherence). Aesthetic Virtues: Pertaining to qualities like elegance and simplicity (e.g., simplicity, unification). Diachronic Virtues: Concerning how a theory performs and changes over time (e.g., durability, fruitfulness). This structured taxonomy offers a comprehensive set of standards for evaluating theories. The paper argues that Williamson’s abduction and Keas’s classification are mutually supportive. Keas articulates the standards implicit in Williamson's "best" explanation. For instance, Williamson's "simplicity" and "elegance" map to Keas's aesthetic virtues; Williamson's "strength" or "unificatory capacity" are expanded by Keas's unification and aspects of explanatory depth. This combined understanding (philosophical thinking as abductive activity, aiming for virtues detailed by Keas) is posited as something that can, in principle, be implemented in an LLM. 4. The Architecture of DeepSeek-Prover-V2 (DSP-v2): An Analogous Case Study To assess the feasibility of an LLM generating good quality philosophy, this section introduces DeepSeek-Prover-V2 (DSP-v2). DSP-v2 is not presented merely as a recent LLM advancement, but as an analogous case study for several specific reasons: Demonstrating Capacity for Complex Reasoning: DSP-v2's success in formal mathematical theorem proving shows that current LLMs can perform complex, goal-directed, and highly structured reasoning. This establishes a baseline capability relevant to the demands of philosophical argumentation. Illustrating Operationalisable Abduction-like Processes: The way DSP-v2 generates high-level proof sketches and then decomposes these into verifiable sub-steps provides a model for how abductive-like reasoning (forming a hypothesis and then testing its components) can be implemented in an LLM. Showing Implicit "Virtue-Seeking" in a Formal Domain: DSP-v2's training, particularly its use of a consistency reward, implicitly guides it towards outputs that exhibit formal analogues of theoretical virtues, such as internal coherence with an initial plan and logical consistency. This suggests that LLMs can be oriented towards desirable output characteristics beyond mere statistical likelihood. Providing an Architectural and Conceptual Precedent: The specific architecture of DSP-v2 (e.g., hierarchical decomposition, feedback loops based on verification) offers a concrete starting point for conceptualising PhiloSeeker. It makes the proposal for adapting such principles to the less formal domain of philosophy appear as a plausible extension of existing AI methodologies. DSP-v2 Overview: An open-source LLM for formal theorem proving in Lean 4 (Ren et al. 2025). Objective: Generate formal mathematical proofs, linking informal intuition to formal verification. Core Strategy: Hierarchical decomposition, guided by informal reasoning. A general LLM (DeepSeek-V3) generates a natural language ‘proof sketch’. This sketch acts as a blueprint for decomposing the problem into formal subgoals (lemmas). Proof Generation Process: The sketch is translated into formal Lean 4 statements (e.g., have clauses with sorry placeholders). Subgoals are tackled individually by a specialised prover model (e.g., a 7B parameter model mentioned in their research). Successful sub-proofs are synthesised into a complete, verified proof. Learning and Refinement Mechanisms: ‘Cold-start’ reasoning data: Pairs successful informal sketches with their verified formal proofs for training. Reinforcement Learning (RL): Primary reward: Binary (correct/incorrect Lean verification). ‘Consistency reward’: Encourages structural alignment between the final proof and the initial sketch. Expert Iteration: Incorporates successfully generated and verified proofs into the supervised fine-tuning (SFT) dataset. Operational Modes: ‘non-CoT’ (non-Chain-of-Thought) mode: For rapid, concise formal proof generation. ‘CoT’ (Chain-of-Thought) mode: Articulates intermediate reasoning steps in natural language for transparency. These architectural and learning features of DSP-v2 provide a concrete foundation for considering how similar principles might be adapted for the distinct challenges of philosophical reasoning, as explored in the following sections. 5. DSP-v2: Operationalising Abduction and Implicitly Seeking Theoretical Virtues This section analyses DSP-v2's operational processes, as described in Section 4, through the conceptual lens of abduction and theoretical virtues (from Section 3). The aim is to demonstrate how its mechanisms can be interpreted as formal analogues of these philosophical concepts. DSP-v2's Core Methodology as an Analogue to Abductive Process: The generation of an initial informal proof sketch by DeepSeek-V3 is analogous to the abductive step of formulating a high-level hypothesis or potential explanation (i.e., "This theorem is true, and this sequence of major steps offers a plausible way to demonstrate its truth"). The subsequent decomposition of this sketch into formal subgoals, and the systematic attempt to prove each one, mirrors the elaboration and rigorous testing of an initial abductive hypothesis. Each successfully proven subgoal lends support to the overarching sketch. A fully verified proof, therefore, confirms the initial sketch as a "good" explanation or a successful strategic outline within the formal system. Through its training (RL and expert iteration), DSP-v2 learns to generate sketches that are more likely to lead to successful proofs, akin to an abductive reasoner learning to propose better, more tractable hypotheses. DSP-v2's Architecture and Learning Mechanisms as Implicitly Seeking Formal Analogues of Theoretical Virtues: Evidential Accuracy and Internal Consistency (Formal Analogues): The Lean 4 proof assistant acts as an objective arbiter. Any proof verified by DSP-v2 is, by definition, formally correct (analogous to ‘evidential accuracy’ relative to the axioms of the formal system) and internally consistent (containing no detectable logical contradictions). These are non-negotiable requirements of its operational environment. Internal Coherence (Formal Analogue): The ‘consistency reward’ in DSP-v2’s RL phase directly promotes a formal counterpart to this virtue. By rewarding proofs that maintain structural alignment with the initial lemma decomposition, the system is encouraged to produce outputs where sub-proofs fit together cohesively and follow the strategic plan. This is analogous to internal coherence, which values the integrated and non-ad hoc arrangement of a theory's components. Simplicity (Formal Analogue): The ‘non-CoT’ mode, optimised for concise formal proofs, aims for a form of syntactic or structural simplicity—achieving the deductive goal with less complex formal output. This parallels the theoretical virtue of simplicity. Clarity and Explanatory Depth (Formal Analogues): DSP-v2’s ‘CoT’ mode, by explicitly articulating intermediate reasoning steps in natural language alongside formal code, enhances transparency and understandability. This provides a form of ‘explanatory depth’ by revealing the inferential structure and rationale, analogous to the philosophical desire for clear and insightful explanations. Learning from "Virtuous" Examples: Through its training on "cold-start" data (its own successfully verified proofs, which are inherently accurate, consistent, and often coherent due to the consistency reward), DSP-v2 learns from exemplars that already possess these desirable formal characteristics. In essence, while DSP-v2 does not engage in philosophical deliberation, its design illustrates how an LLM can be structured to navigate a complex reasoning space using abduction-like strategies and be guided by feedback mechanisms that implicitly select for outputs possessing characteristics analogous to theoretical virtues. This analysis provides the crucial groundwork for proposing that similar principles can be adapted for the philosophical domain in the PhiloSeeker model. 6. PhiloSeeker: A Hypothetical Adaptation of DSP-v2 for Virtuous Philosophy This section proposes 'PhiloSeeker,' a hypothetical LLM adapted from DSP-v2's principles to aim for theoretically virtuous philosophical outputs. Objective and Domain: Primary Objective: Generate philosophical texts (theories, arguments, conceptual analyses, responses to problems) that are cogent, internally consistent, and exhibit high theoretical virtue (as per Keas, 2018). Operational Domain: Natural language philosophical discourse, with its nuances and lack of formal verifiability (unlike Lean 4). Core Strategy (adapting DSP-v2's hierarchical decomposition): A sophisticated general LLM, ‘PhiloSketcher’ (analogous to DeepSeek-V3), would produce an initial high-level ‘theory sketch’ or ‘argumentative blueprint’ in response to a philosophical problem/prompt. This blueprint would articulate main claims, define key concepts, outline sub-argument structures, and potentially flag objections or areas needing clarification, decomposing the task. Process of Philosophical Construction (mirroring DSP-v2's sketch-to-subgoal translation): ‘PhiloSketcher’ translates its blueprint into a structured outline (assertions, questions, tasks with placeholders for elaboration). These individual philosophical sub-tasks (e.g., "provide three arguments for claim X," "define 'phenomenal consciousness'," "formulate response to Gettier problem") would be processed by a more specialised ‘Philosophical Argumentation Model’ (analogous to DSP-v2’s smaller prover model). Adapted Learning and Refinement Mechanisms: ‘Synthetic data for a philosophical cold start’: When PhiloSeeker successfully constructs a complete philosophical theory/argument deemed highly virtuous by evaluators (human or AI), this entire output is paired with the initial argumentative blueprint. This creates training examples linking high-level philosophical strategy with detailed, virtuous execution in natural language. Reinforcement Learning (RL): Central to development, but with a different reward structure than DSP-v2's binary reward. Guided by a composite ‘virtue score’ derived from evaluating the text against Keas’s theoretical virtues (via automated metrics and human feedback, detailed in §7). A critical component: a ‘blueprint-consistency reward’ (paralleling DSP-v2) to incentivise the output to faithfully instantiate the initial sketch, promoting internal coherence. Expert Iteration (adapted): PhiloSeeker learns from its own most ‘virtuous’ outputs (those receiving high ratings from human evaluators or a virtue-discriminator model). These are incorporated back into its training set. Dual Generation Modes (similar to DSP-v2): ‘non-CoT’ mode: Might produce concise, polished philosophical essays or arguments. ‘CoT’ mode: Would explicitly articulate the underlying argumentative blueprint, rationale, definitions, argument development, counter-argument consideration, and connections between theory parts, enhancing transparency and scrutiny. The section suggests DSP-v2's core architectural loop (sketch, decompose, solve, synthesise, refine) is robustly applicable to philosophy. The primary adaptation lies in the nature of the ‘solved state’ (philosophically virtuous output) and the evaluation/reward mechanisms. 7. Training and Evaluating PhiloSeeker This section details methods for training and evaluating PhiloSeeker, emphasising operationalising theoretical virtues (Keas 2018) as the guiding principle. Core Training Loop Adaptations from DSP-v2: Initial Data Seeding: A substantial corpus of high-quality philosophical texts, selected for philosophical significance and exemplification of virtues (clarity, rigour, coherence, etc.). Corpus might require annotation (human or LLM-driven) to highlight argumentative structures, definitions, or virtue instantiations for initial supervised fine-tuning of ‘PhiloSketcher’ and ‘Philosophical Argumentation Model’. ‘Cold-start’ Data Generation (Philosophical): When PhiloSeeker produces a complete philosophical work (essay, theory fragment) from a blueprint, and it is deemed highly virtuous by the evaluation framework, this blueprint-to-output sequence becomes a training instance. Curriculum Learning: ‘PhiloSketcher’ could generate numerous smaller, focused philosophical tasks (e.g., "define 'free will' compatibly with determinism," "formulate three objections to utilitarianism"). These provide denser, targeted training signals for the ‘Philosophical Argumentation Model’. Expert Iteration (Philosopher-in-the-loop refinement): PhiloSeeker learns from its own generated philosophical outputs that achieve a high ‘virtue score’ (determined by the evaluation process). These are added back to the training set. Reinforcement Learning (RL) with a ‘Virtue Score’: This is the most critical adaptation from DSP-v2's binary reward. The score is multi-faceted, derived from Keas’s taxonomy. Operationalising Evidential Virtues: Rewards for consistency with specified empirical data/scientific theories (if prompt demands), or accurate representation/citation of texts. Judging ‘causal adequacy’ or ‘explanatory depth’ of novel claims would likely initially require significant human feedback to train a discriminator. Operationalising Coherential Virtues: NLP tools for detecting overt logical contradictions or inconsistent terminology (for ‘internal consistency’). The ‘blueprint-alignment reward’ strongly promotes ‘internal coherence’. ‘Universal coherence’ could be partially checked against a curated knowledge base. Operationalising Aesthetic Virtues: ‘Beauty’ would likely rely on human ratings to train a discriminator. ‘Simplicity’ might use proxy metrics (ontological parsimony via entity counts; conceptual parsimony via undefined primitive terms; syntactic simplicity via conciseness relative to scope). ‘Unification’ rewarded for connecting disparate problems or deriving many consequences from few principles. Operationalising Diachronic Virtues: ‘Durability’ tested by subjecting theories to ‘objection-bot’ LLMs. ‘Fruitfulness’ might be rewarded for theories identifying unresolved questions or suggesting novel lines of inquiry. Balancing these virtues would likely require multi-objective RL with careful weighting. Hybrid Evaluation Framework (in absence of formal checker): Automated proxy metrics: First-pass signal for certain virtues (consistency, some simplicity forms). Human expert evaluation: Indispensable for qualitative virtues (explanatory depth, beauty, fruitfulness, plausibility). Panels of philosophers with diverse specialisations rating outputs on virtue dimensions using clear rubrics and inter-rater reliability checks. RLHF (Reinforcement Learning from Human Feedback) techniques: Iterative sampling and focused feedback. ‘Virtue-discriminator’ LLM: Trained on human ratings to serve as a scalable proxy for direct human judgment, guiding RL and flagging outputs for human review, with periodic recalibration. 8. Potential Objections and Pluralism This section considers potential objections to the PhiloSeeker project, as outlined in the original draft. Objection: Need for a unique human 'creative spark': The paper's proposal of a system generating complex, virtuous philosophical content implicitly challenges this. Objection: Risk of homogenising philosophical thought: The draft does not detail specific safeguards in this section, but the broader emphasis on diverse virtues and training data might be relevant. Objection: Subjectivity of aesthetic virtues: The paper proposes addressing this by using human ratings to train discriminator models for these virtues. Safeguard Mentioned: Universal coherence with empirical science: This suggests a grounding mechanism to ensure generated theories align with established knowledge where applicable. 9. Summary of Architectural Adaptations This section summarises the key architectural changes proposed for transitioning an LLM from mathematical proof generation to philosophical theory generation. Core LLM Architecture Largely Retained: GPT-style models. Hierarchical prompting strategies. Reinforcement learning (RL) infrastructure. Critical Swap: The Lean-verification filter (used in DSP-v2) is replaced with a 'philosophical virtue-evaluation module.' Additions for the New Evaluation Module: A virtue-discriminator LLM (trained on human judgments of philosophical virtue). A human rating pipeline. Potentially 'objection-bots' to test theory durability. Retuned RL Rewards: Shift from binary correctness to multi-objective optimisation for Keas's theoretical virtues. Maintenance of a blueprint-consistency reward to ensure coherence with the initial plan. 10. Conclusion The paper reiterates its main argument: few fundamental architectural changes are needed to transition from mathematical proof generation to philosophical theory generation. Primary Adaptation: The replacement of a formal verifier with a complex system for evaluating philosophical virtue. This evaluation system is central to the PhiloSeeker concept. Challenges of Philosophical Language: The paper acknowledges that the nuances, ambiguities, and lack of formal verifiability in philosophical language present challenges. However, these are framed as primarily affecting training data creation and evaluation mechanisms, not necessarily posing insurmountable obstacles for the core LLM architecture. Prospect of 'PhiloSeeker': Presented as a conceptually coherent and challenging, yet potentially near-term, research project. Broader Contribution: The endeavour to build such a system would not only advance AI capabilities but also contribute to a deeper and more operational understanding of philosophical virtue itself, by necessitating the clear articulation and implementation of criteria for what constitutes good philosophical work. ## transcript ** It's not a paper, it's still halfway between a draught and a plan. I mean, it's struck there's a lot of stuff in there and it makes sense. So, anyway, read more on yourself. That's one of these and maybe then have another one with more. The other, I want to know what you think of the ideas first, because it's very different arguments. I love it. Just to mention, it's not that important to go to. Every I've probably overdone the detail here, as long as you get the idea. Uh, theoretical virtues are. Okay, do we read everyone? By the way, if we're going this view to Direction it may be also what discussing this with Michelle our colleague because he's working precising that an epistemic virtues. Oh, really? Okay, which one is Michelle who Michelle? Yeah. Yeah he works with. Uh yeah. Yeah. Does moral epistemology, indeed? It is an epistemologist just for for weird Italian Recruitment reasons, it became a moral epistemologist, but interesting. His main mosquitoes and epistemology. You know, the lion on the, the Psychedelics guy, oh, that guy come to the conference in November so no, no. Uh, okay, this sounds cool. It's just thinking, but I don't know, maybe so much to, to even try to to to have also calcium back as so, you have the general methodology, which is on, and then Keys enable us to articulate this methodology. This abductive methodology. And maybe Carson enable us to, to explain what's the results of this methodology. The understanding paper, we, we had at the beginning and now it seems is no longer. That's an okay, right? You're right. It's a political philosopher. Yeah. It's another person. Yeah, just to be sure because otherwise, this one, right? Yeah, yeah. I was just being sure because normally it's me, Give us a very general strategy but we need something more precise and Give us a way of, but at the end you say, oh, this combined understanding, we suppose it as something that can be implemented. So it seems that Williamson plus keys. Give us. So maybe in this way and plus sales. So we are even more Overarching. Metaphilos official account. Also because certain notion seems also to to reappear for his accuracy. Oh certainly yes or yeah, Universal coherence, seems closer to compensiveness. Absolutely absolutely comprehensive said, yeah, now this is a set of standards, it's different but Credential. Vr2 seems to. Guys could I should be accused it fits with other knowledge, this seems much closer to what they mean by which means oh comprehensive but yeah. Yeah. Again, for that next section, there may be too much technical detail there but maybe just so you can get an idea. As to how this because it's just a description of how this llm I like this idea of focusing just on one llm in this way, which is not just general planes and say, oh, there is really something working that way. Oh good, good, I was really worried about this. And also badly that llm was only released two three weeks ago as well. So hopefully nobody else is frightening about this just yet. Yeah, no. I think it's because people say, oh, baby, are you sure are we really like that or was? It seems more just general? In this case, you know. No, no there is a product that is something like that. Does. Uh, the mathematical one. Yeah. What uh, Take a theorem and try to prove it or you also created. Um, sometimes I've got the paper in front of me and I would I guess that is just proving it was usually mathematics. But I don't, you know, maybe he's also generating the full. Yeah, formal theorem holding. Yeah, so yes, the user say oh please find the proof. These are the actions. This is the system. There is this theorem to be proved. Please find the proof and the system and give the proof but then There is this analogy because at a certain point, since that we say that, The system is also meant to to not only find the proper to to, to find the theorem. But maybe let me try and see whether I can find a this passenger, okay? And, you know, this is I'm not an expert on this stuff either. Why we can I think we just have to use it the property as an analogy. Yeah, of course. Okay, now it makes sense. So when you say this theorem is So this is just a, it's not the point, finding the theorem the theorem is just a claim. And the crucial point is this sequence of major stats of a possible way to demonstrate this. So okay, it's okay. Yeah, maybe that's that's the point where I'm not happy. Generally sketches that are more likely to lead to successful proofs, according to an abductive reason and learning to propose better. More tractable hypothesis, that's not correct. Okay? Because proof, I think in the the right analogy is say in maths, we have theorems. Proves. Whereas in philosophy, we have hypotheses. To be supported by arguments. So the analogous of the proof is not hypothesis. The arguement. Okay, let me get this done. I understand you, make sure I forget it, so, yeah. How did you put that theorem to what was the video used theorem to be proved by proofs? Correspond to hypothesis to be. Supported by arguments. I think that that that's the analogy, we we should be relying, okay? And so here, I think people track. Better more tractable arguments, I would say, okay, Then sure in a sense is but that's maybe something more. Maybe this is more specific of philosophy. So, the idea is that, well, you have a possible hypothesis. You beat an arguement and you see what the arguement is. Is it is persuading. Then you have another hypothesis and you build another arguement. And then by by selecting the arguement, you can also select the hypothesis or that's, that's something that mathematics seems less common people don't just say oh we have a theorem we have another theorem and approved because the proof is just yes or no. Whereas in Philadelphia. Yes but yeah, maybe that's something. Okay, we can develop but I think that at this level, the crucial analogies between producing the arguments. And so the just like, improves are meant to prove clearance arguments are meant. You are good for Um, let me know when you finish that section as well. Because again there's a lot of information here which can probably be I mean this is this is where you are more or less. Yeah so you would sort of talk about how we can think of this as implicitly doing abductive stuff towards virtues, right? Um, All this is quite interesting. It gets very when I start actually suggesting how to make the thing in six. It's quite speculative and I was wondering, maybe we could talk a little bit before you look at that. Anyway. Okay. Yeah I'm just going to finish fighting. Yeah. The, the main concern I I have now, at the end of five. Yeah. Is that since a bit of called this describe the, the the proving system in terms of abduction, because proving is is a paradigmatic example of deduction. Studio Williams. Say oh also mathematics founded. Yeah, yeah. Okay. But when it comes to proving proving is really. Yeah. Yeah. Hmm. There's a good point. You put it like that, I Is it that? So the easiest fix with the easy, the easy fix would be. To change five quite significantly or even remove it. So just say actually no no DSP version 2 does not abduct may. I yeah maybe even more straightforward even though I don't feel I'm not sure that it's this fair. But one way to to ride would you just say okay the the approver has deductive virtues but we can find the abductive counterpart. That's that's what that's the easy way of doing it, right. Yeah. So maybe even actually just outside, this is deductive, which actually have this in terms of deductive virtues, perhaps. Yeah. And then yeah. Then. I know we're going to try and rearrange the, the trading methods and the mechanisms. Yeah. Because yeah, because that's nice because then it really does just hinge. It seems to me that so that's actually a really nice way of thinking about it actually. I think the two things that this hinges on numbers, the, the idea of having the philosophy llm would be, can we operate operationalise, abduction, generally. And worries which you've kind of talked about last time as well. There's no equivalent to the Lem whatever they call it with the Lenny's. The Lemn is the prover in the maths deepsea and the thing. So, Lemma sorry. Um, As you said last time, Matt's there was a definite right and wrong. Okay, this is correct. This is not and they they control for this inside their model. They say there's like a proof chequer that they're plugging into, right? So that's that's the other big problem. I think this paper would have to say, which is to say we don't need such a strict thing here but we can still get. Yeah. Yeah. These are different. Um, in one way, in one case, we have a more deductive process, whereas inferior. So this is I think that the general schema remains that. So there is a correspondence. But students are not hypotheses and proofs are not arguments because proof of deductive. They have they can be checked. They are matter of yes or no. Perfectly sure that it's a good. It's a it's a sound Groove or just a ground progressing philosophy. It's always more about Yeah, contending and more mature, probability of some. But we may say, okay, despite this, this analogy, there is a structural analogy. And there are virtues that theoretical beer to six theoretical is the overall category that encompassed. But induction and abduction. There are theoretical vehicles that are both. So from these, we can have a sort of Version which is this but it's not a problem because philosophy is less perfect than mathematics. So the the the philosophical proverb would just reflect the features of philosophers compared to Google philosophy as compared to good mathematics. It's just that the same virtues are to be declined differently. Yeah. Way, it gives us more. More work to do in the more constructive work as well. Okay, we have something in mathematics and we are now philosophy. Yeah, yeah, yeah. Okay, good. So can I now start six to you? You can have a look at six is basically, just a very, a sort of me, talking to the llm and saying, how can I transfer Um the how can I term approver into the philosophy care? Whatever it's called. So have a look. I'm going to run to the bathroom. Yep. Another question. Because it's true. That in a sense, it's the the objective. If this philosophy would be to say, generate philosophical test theory, arguments conceptual analysis response to problems. On the other hand, if you want to keep the analogy, maybe Just be better just to have arguments. Just saying it's true. It's not a philosopher because philosophers don't work this way. If he doesn't really not saying oh please. This is the this is physical, is with an arguement, some philosophers. Maybe work like that. Oh, and I've realism this giant TV is I make an arguement for that. So In a sense, it would be more accurate to philosophy. What you what you what is written here? The objective is to generate philosophical test series, but if we want to, to keep the energy the analogy with maths, uh, At least at the structural level. Since we have already some this analogy in terms of deductions, then maybe keep it because. Yeah, because they're not, they they haven't created, provo to write mathematical papers, right? They've created prove it. Yeah, just seems that Google the way which approval work is. Oh, this is a theorem, please find the proof. Uh yeah. This is a philosophical statement, please find an arguement something like that. Yeah, I want to spend any work specifically and what counts as an arguement in philosophy. I know you can just say premise premise conclusion, but I wonder if anyone's thought about it harder. I don't know. Yeah, another problem is that There are already arguments whereas when you when you look for a theorem, there are no proofs on the market. They're not competing proofs because if there is a proof that you don't need to another well maybe you can find a better proof but that's not not what mathematicians are care so much about they when they approve their super happy and they can just move to the next to the next tier. Then uh, whereas with philosophy is different because you are. So, the system should not just find an arguement because finding the arguement is not winning. What is much if you find that approved, you, you've won. Yeah. But oh, it's okay. The arguement is already on the market. So that's another thing that. Yeah, I think we have to think about because that's what motivates maybe having a primary objective but not just Halfway between an arguement and a full paper or something like that. So, Well, maybe we can just put this in the In the paper problematize. Take these problems as matter to be discussed. Because maybe. Yeah. Just Yeah, the problem of existing arguments which is it seems as strong as analogy because The Proverb doesn't have to to deal with existing proofs for the reason. I I just said But it seems that that And philosophical de la land. Instead, it's meant to to deal with. Pre-Existing arguments. So it's also in the philosophy, often say this, it's not just a matter of writing. Arguement it's also a matter of evaluating arguments, assessing arguments. Yeah, criticising arguments. For as much as far as I know, It's not about, maybe it may happen but it seems more like accidental. Yeah, it's not the core because it's it's not too difficult, it's not, it's controversial with an arguement day, he's working or not working. So if it was On otherwise, the arguement is just, the proof. Is just not, it's not even a proof. A proof that doesn't work, is not a proof. Yeah. But as an arguement, that it's perfect. Good. Yeah. Hmm. But I don't think it's, I don't think if you said that, yeah, it's It's too big a problem. It's a problem, but it's not too big. I think no, no, not destroying the the strategy. Is maybe it's Yeah, this philosophical proverbs should should rely on the analogy. Yeah, to to generate Different arguments and maybe compare them. I don't know. Yeah, but that, that seems something more than the matter require a different. Maybe a work closer to that of traditional lamb rather than this adult lamb. This is the But maybe, maybe we can just instead having something. This will be super useful. Nevertheless saying, no, maybe no. We just have Driving driving on the analogy with the fluver we have Yeah, we have a way of creating an arguement Yeah. Now that yeah. Okay my problem is it seems to be this one that it seems that making an arguement in philosophy. Is also a matter of engaging with other arguments. Well, yeah, we're still Standalone almost, right. Yeah. On the same. Anyway, let me see what follows. I mean, again it's just a lot of sort of technical ideas about how we could how it could be. Uh, training up for example engineered. Yes, this is correct. So produce high level Theory sketch, argumentative blueprint I would say more again. I I In response to a philosophical problem. Prompt. To me, if If there is the analogy. Again, if we say produce and an initial eye level, arguement sketch or arguement group in response to a philosophical statement, Otherwise, but I don't know, maybe I'm wrong. My my sense is that the analogy breaks if we, if you replace problem philosophical problem And we're going to place arguement with T or E. Then the two things quite different because it's no longer a matter of of steps. And the chain of reasoning is more a matter of exploring and logical space. So we have we get closer to the initial idea of understanding and death and which seems a bit intentional with the, the arguement idea, I think that's the problem. We we face it from the beginning. The idea of philosophy is understanding suppose to philosophymes arguement which are connected. But in a sense, I'm not I'm not completely identical because students will understand this thing. Oh let's say then you can use arguments, you say oh maybe let's see where the digital relations also or not and but then you have you need one arguement for each uh say lying in the network. So understanding it seems more a long long period goal whereas arguement is more a short period gone and Commendiano. I, I Every sort of. Uh, say language problem, because I I've noticed that, um, everybody in the Japanese group, they call takuya nikawa okay. So because I I remember that I've read but I I know in almost 10 years and in Paris, everybody is friends. They everybody calls him takuya. Yeah, what's his name? What's your name? Yeah. It's just a weird thing in Japan. They have maybe the, the Japanese families are more formal. Yeah, it seems strange because since it's changed. Yeah. So maybe Uriah and the others are working pretty mistaken. They keep and it's just been too kind. And is this not surprising because it's a very kind thing. Yeah. I remember at a certain point in Paris, there was a discussion whether the right way to call him was takuya or takuya, but it seems that that there is a deeper problem. It's not just a problem going to place the accent, it's what is that? Which is the name, which is the sunny Provide three arguments for clinics design, political systems, forward response to get their get the problem. Provide to us. And the first one may work because it's similar to the, so the structure of the blueberry but then Define, then yeah, it's at least from the yeah. From the mechanisms described, it's not obvious. How that's going to be done. Yeah. Yes, yes, this is quite This is yeah, clearly it needs to be tightened up. Maybe we can just focus on provide. An arguement for claim X that would be already, upgrade Africa achievement. Then we may say in the last part of the paper maybe we can say okay, should philosophy is not just Eggs. Please write an arguement. Yeah yeah. Maybe the at least at this level we can have a sort of division of labour. So the human look for an interesting hypothesis and then ask Llm, please write an arguement. I feel bad. You're reading all this detail there. I hope it's not too boring. I just put the detail down just to find. Yeah, but it's in this way. This schema way, it's easier to navigate than just having a compact. Maybe this will also give us. Material for two papers. So one paper, just for the arguement, which is this one, and then we can go back to the initial idea of understanding and saying, oh now we have a theory of how to build an arguement. Let's see how we can use it and then also for building theories and not just arguments. Cool. Yeah. Great Now, the main concern is whether there is be an analogy. Between deduction and abduction. Yep. The problem philosophical arguments is more. Defending premises. Usually the, the the deductive part is trivial. Yeah. Yeah. Because yeah, you you start your paper win. Here is my arguement. I will now be if you, if you have Ask the basic. Principle of logic. It's quite rare. That philosophy can make mistake or the objection are about how. Yeah. Yeah. Whereas in maths, that's all the key. The key of all that is just it's not, that's all this premise. Doesn't always premise. Always all the maths, the more, it's more. Enchainment of the passages. So, this is this analogy that, and that's why an abduction in philosophy, seems to play a lot because It's a way of supporting the premises is connecting the premises. But then seems that that what is crucial to good philosophical. Arguments seems pretty different from what is crucial to a good? Mathematical proofs. Yeah. And so this may threaten the The idea that We can just here just uh adapt adapt approver to have. Yeah. Um, I mean, I think what what needs to happen now is it needs to go away and learn more about Abduction in philosophy and deduction in mathematics and stuff like that. Um, I mean, I, I understand that the issues here but I think I'm hoping It seems obvious. It doesn't seem obvious that irresolvable I'm just honestly please you like the idea because it's so far away from what we did last time. Yeah, I think it's another paper. It's more about arguement and how to, to use a approval to Something have you seen Google score or on the there should be something I think in the analogy between arguments and proofs philosophical arguement mathematical proofs. They seem to have interesting correspondences. I'm starting. It's also available. It's true that you just need. I think the key of our study is this drawing the analogy between the, the the epistemic virtues of the proverb and the virtues. These are not the same virtues, but maybe we can just say find Of analogy. Is that a reason for thinking they could be operationalised in the same way, right? Exactly. And screw the two instances. This is no no. Should any prover is formally correct? Analogous to evidential accuracy and internally consistent But yeah, internally because formally correct. Internet consists and for approved seems a little bit more or less. I would say the same thing. Uh, so But maybe that, that's what we want. We will just want to another of who receive Yeah, this is the the pump you need. That what the analogy is to breaks, is that In in maps, it just you Not not having contradictions. Having consistency is all that matters. Yeah, having a good proof whereas in in philosophy. The point is also. But it's also. The strength of of the premises that are that are used in the In. So, it's not just the, the logical consistence between the premises is also whether the premise is Uh, uh, Are supported or not by evidence? Yeah. There's not this problem premises. The premises have just generated in turn and what will support them. It's just a logical coherence. Yeah. So, that's So it seems that to, to support the premises, which seems the key operation in our philosophical. Arguement the the the the system is meant to to do something. Of a different kind. Than than the prover, a small man to to Okay, look for for He has to look outside, also the system of Axioms or unless a system of action, we mean just Or philosophical knowledge over all non-available knowledge. Maybe that that's maybe another Because yeah, yeah. Maybe this is the other interesting point. That proof means drawing conclusions. Finding chains of reasoning from a quite limited set of actions. That's what make months in a sense easier than philosophy. There's no there's no grey area, right, there's no fuzziness. The principles are there. The the the, the, the, the, the the gain is just just the chest? The rules are those, then the moves are just fixed by by the the rules. Whereas which are the actions of philosophy and that's that's sort of yeah, seems to be unless you you have Espinoza then because you want to do philosophies mass or it beats his own axioms and then you draw conclusions. Exceptionally, very philosophy. They are axioms or they build their own say, oh, maybe this is I assume that I assume that. So again. That doesn't break the energy, but it has to be problematized to see the same things that referees may may obviously. Yeah. Right, say, okay. But but What, what's? And prove. Just walk by by relying on these premises. We were. And then saw the, the task of the proverb is to, to given the given, the, the axioms given the Theory and find the connection that enable one to obtain the current from the axioms. Yeah, Yeah, so we can give that to the the philosophical. Of the children. So the hypothesis. But reach out the analogies of the actions. Yeah. I thought some interesting stuff. Methodology is maybe interesting to look into. What do you think? Is that a blank? It's Oxford. University Quest book. I'm just very rough feelings that the the the previous version, the death and bass version was more manageable in the sense that it seems to to individuate the features of specific of philosophy. And which is also closer to abduction in English and also to that, but Yeah, the problem is whether this, whether the the proverb is the right, uh, way to go because there's something I just couldn't. Get any further with it. Yeah, I just what I found. I liked it a lot in the first version of the Descent bass. The paper is that since that this idea of a network from dependency, seems to, to have an interesting analogy with the latent space. Absolutely. This seems something can be to me, more convincing, I see more analogy between the the, as far as I can understand. Both on the latent space is the way in which Concepts are organised in the, and this idea of network of dependencies because it seems with this network of dependencies can be created in a statistical, uh, manner. And that seems to do even to display, why even general purpose? Seems to be good in doing something in philosophy, do something this and think philosophy, this is more in a sense. It's more promising because it's even So, attitude manager, and those maybe require also knowledge in epistemology That we like, and they require a lot of time to acquire Make me a bit less Optimist. In a sense that There isn't an interesting analogy which is the one we will just discuss because it's true that that here are to to proof in maths. What hypotheses have to arguments in philosophy. But, Yeah, there are also this analogy that we just discussed and I wonder what it is. This analogy is I'm gonna sing the ship. Yeah. Give me some time to think about it. Yeah, I'm still uh, I still think I could. Yeah. I I do see the issues here, but I think like, I'd like to try and make it work properly. I'm more enthusiastic about this one than the Dells and stuff at the moment. Okay. I just sort of, I didn't know where to go with the Dells and stuff in the end. Yeah. Um, Um, I mean, Have another meeting early next week, and I'll try and give it maybe in the. Meanwhile, if you can see It would be great to find out. Is I have the sense that there should be somebody who has written something serious on the analysis between proofs and Arguments without the need of reading 500 pages book. On epidemology. Yeah. Just uh yay and while I'm doing some research, yeah. Yeah you can see Google typing away now. Yeah, so this is not finished yet but you see it's 60 websites? So my proof was just has anyone written about the scenarios and differences between yeah, mathematical proofs. And let's see what it comes up with. See, I find the stuff last night because he's just thinking it's you can just see what it's doing. Yeah. I mean, I'll send you this document, come from deeply grain and ways of thinking devices, such as that, our competency must do, isn't just about another, but they're going to do. But this is just about. Yeah, yeah, it's it's working through it, okay. Anyway, and you can I think you can keep this weird. Yeah, yeah. We don't start till two o'clock, right? What time did we start? It's two o'clock. We start this. We start the meeting at two. No. Okay, just as you guys seen the movie. So you Okay. They're not mistaken. They're just uh, yeah. Yeah. De monkey Madre decor. **